Abstract Purpose: To evaluate synthetic images generated from a 3D-QALAS sequence with scan-specific deep-learning-based subspace reconstruction (Zero-DeepSub) relative to conventional MRI sequences in pediatric brain MRI. Materials and Methods: This prospective study included 26 pediatric patients (mean age 8.4 years) who underwent clinical brain MRI between November 2023 and January 2024. Synthetic T1-weigthed, T2-weighted, and FLAIR images were generated from quantitative maps using a 3D-QALAS sequence with Zero-DeepSub reconstruction. Two neuroradiologists independently assessed seven predefined imaging findings on synthetic and conventional images. The reads of a third neuroradiologist were used to adjudicate discrepancies; this neuroradiologist performed an independent semiquantitative image evaluation using a 5-point Likert scale to compare image pairs. Statistical analysis included sensitivity, specificity, PPV, NPV, Cohen’s kappa, and Wilcoxon signed-rank tests. Results: Synthetic images showed high sensitivity and specificity for mass/lesion, encephalomalacia, and collections, with perfect agreement between readers for these findings. Gliosis demonstrated high sensitivity but moderate specificity for one reader. Abnormal enhancement had the lowest sensitivity (0.40). Interobserver agreement was moderate for gliosis (κ = 0.55) and almost perfect (κ = 0.83–1.00) for detecting mass/lesion, encephalomalacia, collections, and abnormal enhancement. Semiquantitative image evaluation showed equivalence between synthetic and conventional images for FLAIR, T1-weighted, and post-contrast sequences (p > 0.1), while conventional T2-weighted images were rated significantly superior (p Conclusion: 3D-QALAS with Zero-DeepSub reconstruction enables the synthesis of high-resolution, clinically interpretable brain images in pediatric patients, including post-contrast sequences. This approach demonstrated strong diagnostic agreement with conventional MRI and holds promise for reducing scan times in pediatric neuroimaging protocols.
Ferraciolli et al. (2025) studied this question.